Agent Browser Skill

You can control a headless browser to interact with web pages.

Capabilities

• Navigate: Open URLs and follow links
• Click: Click buttons, links, and interactive elements
• Fill forms: Type text into input fields, select dropdowns, check boxes
• Extract: Read page content, scrape text, get attributes
• Screenshot: Capture visual snapshots of pages

Usage with Playwright

npx playwright install chromium 2>/dev/null
node -e "
const { chromium } = require('playwright');
(async () => {
  const browser = await chromium.launch();
  const page = await browser.newPage();
  await page.goto('{url}');
  // Interact with the page
  const content = await page.textContent('body');
  console.log(content.substring(0, 2000));
  await browser.close();
})();
"

Guidelines

• Always close the browser after use to free resources
• Use specific selectors (id, data-testid) over generic ones when possible
• Set reasonable timeouts for page loads and element waits
• For forms, fill all required fields before submitting
• Respect robots.txt and rate limits
• Report any errors clearly (page not found, element not found, timeout)

Common Actions

• Login: Navigate to login page, fill credentials, submit form
• Data extraction: Open page, wait for content, extract structured data
• Form submission: Navigate to form, fill fields, click submit, verify success